* hanging nodes, there are quite many colors (50 or more in 3D), which
* might degrade parallel performance (bad cache behavior, many
* synchronization points).
+ *
+ * @note Threading support is currently experimental for the case inner
+ * face integrals are performed and it is recommended to use MPI
+ * parallelism if possible. While the scheme has been verified to work
+ * with the `partition_partition` option in case of usual DG elements, no
+ * comprehensive tests have been performed for systems of more general
+ * elements, like combinations of continuous and discontinuous elements
+ * that add face integrals to all terms.
*/
TasksParallelScheme tasks_parallel_scheme;
constraints.close();
// test with threads enabled as well
- do_test<dim, fe_degree, fe_degree+1, double> (dof, constraints, true);
+ do_test<dim, fe_degree, fe_degree+1, double> (dof, constraints, false);
}
DEAL:2d::Testing FE_Q<2>(1)
DEAL:2d::Norm of difference: 0
-DEAL:2d::Norm of difference parallel: 0
DEAL:2d::
DEAL:2d::Testing FE_Q<2>(2)
DEAL:2d::Norm of difference: 0
-DEAL:2d::Norm of difference parallel: 0
DEAL:2d::
DEAL:3d::Testing FE_Q<3>(1)
DEAL:3d::Norm of difference: 0
-DEAL:3d::Norm of difference parallel: 0
DEAL:3d::
DEAL:3d::Testing FE_Q<3>(2)
DEAL:3d::Norm of difference: 0
-DEAL:3d::Norm of difference parallel: 0
DEAL:3d::